a

r
X
i
v
:
1
2
0
4
.
0
2
3
1
v
1


[
c
o
n
d
-
m
a
t
.
m
t
r
l
-
s
c
i
]


2

A
p
r

2
0
1
2
Magnetic superlens-enhanced inductive coupling for wireless power transfer
Da Huang,
1
Yaroslav Urzhumov,
1
David R. Smith,
1, a)
Koon Hoo Teo,
2
and Jinyun
Zhang
2
1)
Center for Metamaterials and Integrated Plasmonics, Duke University, Durham,
North Carolina 27708, USA
2)
Mitsubishi Electric Research Laboratories, 201 Broadway St. 8
th
oor, Cambridge,
Massachusetts 02139, USA
(Dated: 3 April 2012)
We investigate numerically the use of a negative-permeability perfect lens for en-
hancing wireless power transfer between two current carrying coils. The negative
permeability slab serves to focus the ux generated in the source coil to the receiver
coil, thereby increasing the mutual inductive coupling between the coils. The numer-
ical model is compared with an analytical theory that treats the coils as point dipoles
separated by an innite planar layer of magnetic material [Urzhumov et al., Phys.
Rev. B, 19, 8312 (2011)]. In the limit of vanishingly small radius of the coils, and
large width of the metamaterial slab, the numerical simulations are in excellent agree-
ment with the analytical model. Both the idealized analytical and realistic numerical
models predict similar trends with respect to metamaterial loss and anisotropy. Ap-
plying the numerical models, we further analyze the impact of nite coil size and
nite width of the slab. We nd that, even for these less idealized geometries, the
presence of the magnetic slab greatly enhances the coupling between the two coils,
including cases where signicant loss is present in the slab. We therefore conclude
that the integration of a metamaterial slab into a wireless power transfer system
holds promise for increasing the overall system performance.
a)
Electronic mail: drsmith@ee.duke.edu
1
I. INTRODUCTION
Since the earliest suggestions of wireless power transfer (WPT) by Tesla in 1891
1
, the
concept of conveying electromagnetic energy from one location to another without wires has
captured enormous interest
24
. Stimulated by the growing demand in mobile electronic de-
vices and electric vehicles, numerous eorts are now underway to develop new congurations
capable of transporting energy wirelessly, and generally seeking to increase the eciency of
all WPT schemes as much as possible
59
. WPT techniques have crossed over from research
to product development, with several commercial units being oered for charging low-power
consumer
1012
and medical electronics
13
. The majority of contemporary WPT schemes em-
ploy a near-eld based, inductive coupling mechanism to achieve high transfer eciency over
short distances (1 5 cm); for acceptable eciency in these schemes, the distance between
the source and receiver must be considerably smaller than the dimension of the transmit-
ter or receiver
14
. This proximity requirement makes the integration of WPT protocols into
many electronic systems dicult or unfeasible, since the transfer eciency in the near-eld
schemes decreases as a high-exponent power law with the transfer distance. The transfer
eciency is dened here as the fraction of power delivered to the resistive load relative to
the total consumed power.
Recently, a WPT system with a relatively high transfer eciency over a moderate prop-
agation range has been developed, based on a pair of inductively coupled, resonant coils
with large quality factors (Q-factors)
15
. In subsequent experiments performed at Intel, the
eciency in this resonant WPT system was shown to be as high as 75% over a distance of
0.5m between the source and receiver coils
16
. Two key factors in the resonant WPT system
determine the transfer eciency: (1) the Q-factors of the resonators and (2) the mutual cou-
pling strength. Strong coupling between the two resonators increases the energy exchange
rate, thus increasing the power transfer eciency. Though high-Q resonant WPT systems
can have excellent power transfer eciency, they are usually not preferred in practical appli-
cations since the coils tend to be more sensitive to the environment, and dynamic control is
dicult to implement
17
. Nevertheless, given the demand for wireless powering and charging
of devices, the eciency associated with WPT between resonators motivates the continued
exploration and improvement of such systems.
An obvious means of increasing the eciency of an inductive, resonant WPT system
2
is to increase the mutual inductive coupling between the source and the receive resonators.
Because the distance between the source and receiver is so much smaller than the wavelength,
the relevant eld distribution is quasistatic and the inductive coupling relates predominantly
to the amount of magnetic ux emanating from one coil that is captured by the second coil.
Enhancing coupling eciency equates to modifying this eld distribution, which in turn
means focusing or otherwise controlling the near elds.
A convenient means of controlling elds is through the use of a material. High perme-
ability magnetic materials, for example, can guide magnetic ux and are used in motors
and transformers
18,19
. Inherently magnetic materials (i.e., those that derive their magnetic
properties from electron spin), however, tend to be heavy and cumbersome, and have limited
ranges of applicability. Additionally, one has very little control over their intrinsic loss tan-
gents. Over the past decade, the concept of articial magnetismusing inductive elements to
mimic magnetic mediahas been introduced and achieved widespread interest
1921
. In par-
ticular, the use of articially structured metamaterials with negative eective permeability
and permittivity has led to new opportunities for managing near-elds, as exemplied by
the perfect lens. The perfect lens, introduced by Veselago in 1968
22
and further devel-
oped by Pendry et al. in 2000
23
, is a planar slab whose electric permittivity and magnetic
permeability both assume the values of -1 at a given frequency; a source placed on one side
of the perfect lens is reproduced as an image on the other side of the slab, with both far-
and near-elds refocused to the image
2329
.
In the WPT context, the perfect lens geometry is a good starting point for improving the
eciency of coupling between two coils. Viewing the rst coil as a source, any loss in transfer
eciency to the receiver can be explained using the concept of magnetic ux divergence,
which is bound to happen at distances greatly exceeding the dimensions of the source. A
perfect lens which captures the near-elds should be able to refocus the ux at the receiver,
thus enhancing the coupling. The use of a perfect lens to improve WPT eciency was con-
sidered by Mitsubishi Electric Research Lab
30
, who, in recent experiments
7
, demonstrated
eciency gains in a resonant WPT system. A rigorous analysis of the coupling enhancement
oered by a perfect lens situated between two magnetic dipoles was subsequently performed
by Urzhumov et al.
31
. One of the key conclusions of that work was that the eciency
increase through enhanced mutual coupling could prevail over the reduction in eciency
caused by material losses, assuming realistically large resistive loads on the receiver side.
3
This conclusion is critical for the consideration of metamaterials in WPT schemes, since
metamaterials are frequently formed using conducting circuits that can exhibit signicant
Ohmic losses.
The analysis published by Urzhumov et al.
31
employed a simplied geometrysuch as
an innitely large slab and point dipoles for source and receiverto obtain closed-form,
analytical expressions that would provide initial insights into the possibility of eciency
gains with metamaterials. Here, we make use of full-wave, nite-element based simulations
to analyze the mutual coupling between nite-diameter coils situated on either side of a
nite-sized planar lens. We also make use of the above-mentioned prior analytical results to
benchmark our numerical simulations in the limiting case of vanishingly small coils and very
large-width slabs. After conrming the agreement between simulation and theory for several
test cases, we next consider more realistic geometries, investigating the eects of nite coil
diameter, nite slab width, and material loss and anisotropy.
II. THE NUMERICAL MODEL
We analyze the mutual inductance between the coils numerically by evaluating the mag-
netic ux in the receiver coil generated from the source coil in COMSOL Multiphysics. Here
we model the resonant coil by a nite-size coil composed of a single turn of conducting wire,
which forms a magnetic dipole. The magnetic dipole approximation is applicable to this
geometry since the modes corresponding to deeply sub-wavelength, high-Q coil resonators,
such as spirals or solenoids, are predominantly magnetic-dipolar in nature.
In general terms, the eciency of a WPT system consisting of a source and a receiver
coil is
31
=
P
0
2
P
1
+ P
2
=
R
2
R
e
2

1 +
(1)
where
=
P
2
P
1
=
R
e
2

2
|L
21
|
2
R
e
1
|Z
2
|
2
. (2)
Here P
0
2
is the power extracted from the WPT system at the receiver end; P
1
and P
2
represent
the power dissipated at the source and receiver coil, in which the power dissipated by the
load presents as loss, and both the Ohmic loss in the system and the radiation loss from
4
10 20 30 40 50
0.9
0.95
1
1.05
1.1
Frequency (MHz)

D
Is
B

Source coil Receiver Coil

Is
Source coil Receiver Coil
L
W
(a) (b)
(c) (d)
Surface wave
Z
Z
S
y
m
m
e
t
r
y

A
x
i
s
Source
Receiver
Z
Source
L
W
Receiver
Z
FIG. 1. (color online)(a) Two-coil coupling system; (b) Two-coil coupling system with MM lens;
(c) The 3D system conguration with MM slab and its equivalent axisymmetric 2D model in
COMSOL; (d) The ratio between the numerically calculated and analytically predicted mutual
inductance in the conguration shown in Fig.1(a).
coils are included; R
2
is the resistive load at the receiver coil; R
e
2
and R
e
1
are the eective
resistances at the receiver and source coils, respectively (both include the loss from the
self and mutual inductances); is the operating frequency of the system; and Z
2
is the
eective impedance of the receiver coil, including the self-inductance and capacitance. After
having optimized all other factors, Eqs.1 and 2 show that the transfer eciency increases
quadratically with L
21
.
The mutual coupling between the two coils can be evaluated by numerically calculating
the mutual inductance between them, which is the ratio of the induced voltage on the
receiver coil and the excitation current on the source coil, or
L
21
=
V
induced
I
s
=
j

S
2

BndA
I
s
=
j

S
2

Ed

l
I
s
(3)
where

B is the magnetic ux density which can be computed in full wave simulations; n is
the unit vector normal to the surface of S
2
;

E is the electric eld;

l is tangent vector to
the loop of innitesimally thin wire, and S
2
is the area encircled by the receiver coil. The
two-coil system diagram is shown in Fig.1(a).
The presence of a metamaterial lens is simulated by including a slab of material with nite
diameter W and thickness L placed between the two coils, as in Fig.1(b). The dominant
eld components are magnetic, since D
0
, where
0
is the free space wavelength and
5
hence the system operates in quasi-magnetostatic regime.. For this reason, it is expected
that only the magnetic response of the slab will impact the near-eld focusing, and thus
only a magnetic component is assumed in this study. The magnetic slab is also desirable
since it is easier to fabricate a metamaterial where fewer elements of the constitutive tensors
need to be controlled. We consider then magnetic slabs with diagonal material properties,
where the material properties can be represented as
e
= [
z
,
y
,
z
],and
e
is irrelevant.
For simplicity, we assume
e
= 1.
To perform the preliminary comparisons between the numerical and analytical models, an
axisymmetric two-dimensional geometry is simulated, in which the current coil is represented
by an out of plane line current, shown in Fig.1(c). As shown, the axisymmetric geometry
assumes a nite-diameter, cylindrical disk (which we refer to as a slab or lens in this
paper). Note that this rotational symmetry only exists for the z-oriented dipole shown in
Fig.1(a), which is the second case studied in Ref.31. For the other dipole orientations, such
as x-oriented dipole, this rotational symmetry is not present and the presented numerical
model is no longer valid (a full 3D model then can be built for arbitrary dipole orientation
analysis
32
). However, the system shown in Fig.1(a) is an optimal conguration since the
coupling is a factor of two stronger than for the rst dipole orientation case considered in
Ref. 31. The magnetic ux density B can be precisely determined over the surface enclosed
by the receiver coil through a full wave simulation. Thus, we can numerically evaluate the
mutual inductance between the two coils since all the terms in Eq.3 can be determined in
the simulator.
In Ref. 31, the mutual coupling between two magnetic point dipoles is studied analytically
in the presence of a negative permeability slab. The mutual inductance is calculated and
presented as a closed form expression. The resonant coil is rst approximated as a single coil,
nite-sized coil with a constant current and innitesimal coil cross section area, and then
further reduced to a magnetic point dipole with equivalent magnetic dipole moment M =
I
s
A = R
2
I
s
. In the present numerical study, the resonant coils are also approximated by
nite size single coil, but are not further reduced to point dipoles. The analytical expression
for the mutual inductance in the presence of the slab is:
L
21
=

0
R
4
2
4/
x
a(2L)
3

b/a, 3,
L + D L
2L

(4)
where
0
is the free space permeability;
x
,
y
,
z
is the eective permeability of the MM lens;
6
=

x
/
z
; a = (/
x
+1)
2
; b = (/
x
1)
2
; and
L
is the standard Lerch transcendent
function, as dened in Wolfram Mathematica software and Gradshteyn-Ryzhik
33
.
In the absence of the slab when [
x
,
y
,
z
] = [1, 1, 1], the mutual inductance between
the two coils reduces to L
vac
21
=
0
R
4
/2D
3
, when the two coils have the same radius
R. In the following, we will use the enhancement factor dened as the ratio between the
calculated mutual inductance and theoretically predicted mutual inductance in vacuum as
the criterion by which to evaluate the enhancement of mutual coupling that occurs in the
presence of the slab.
=
L
21
L
vac
21
(5)
We rst investigate numerically the retardation eect associated with the mutual induc-
tance between the two coils in free space. The analyzed system consists of two coils, each
with radius R = 0.01 m and separated by a distance D = 0.5 m. the frequency of the
source is swept from 5 MHz to 50 MHz, corresponding to wavelengths between 6 m to 60 m.
At the smallest wavelength in the frequency bandwidth, the radius of the coil is less than
0.2% of wavelength, and the transfer distance is less than 10% of the wavelength; we thus
identify this conguration as a sub-wavelength WPT system. As shown in Fig.1(d), the
mutual coupling calculated from the numerical model is essentially the same as predicted
by the analytical model, deviating by less than 2% of over the broad frequency range. Any
discrepancy between the numerical and analytical models is partially attributable to the
nite radius of the wires, nite diameter of the coils, and also the numerical error from the
applied FEM simulator.
In a practical WPT conguration, the source and receiver(s) are solenoids or spirals of
nite diameter. Also, any metamaterial structure that implements the negative permeability
lens will have nite extent in the transverse direction. In the following section, we investigate
the impact of the nite size of the coils and the slab. If not otherwise specied, the frequency
of operation is 10MHz, consistent with the frequency band used in previous experimental
investigations and some commercial implementations
7,15,16
; the radius of the source and
receiver coils is R = 0.01m; the transfer distance is D = 0.5m; the diameter of the negative
permeability slab is W = 2m and its thickness L = D/2 = 0.25m. Note the thickness
corresponds to the perfect lens condition
23
.
7
III. FINITE SIZE AND LOSS EFFECTS IN ISOTROPIC-PERMEABILITY
LENS
We rst investigate the eect of nite coil size on the mutual inductance in the absence
of the slab. In the analytical study, coils were approximated as point dipoles, while in the
numerical model the coils have nite radius R. To study the impact of nite radius coils,
we vary the radius in the simulation, starting from a very small radius (relative to the
wavelength). As a function of increasing radius, the computed mutual inductance deviates
from the analytical model, as expected. As shown in Fig.2(a), when the coil size reaches
roughly 0.001
0
, where
0
is the free space wavelength corresponding to the source frequency
(10 MHz), the dipole approximation used in the analytical calculations is no longer accurate.
The comparison in Fig.2(a) shows that the deviation in the mutual inductance between
two nite size coils and two idealized dipole remains within 15% when the coil radius is
approximately 0.4% of the free space wavelength, or a quarter of the transfer distance.
Further increasing the coil radius decreases the mutual inductance to the point that Eq.3
no longer provides an accurate estimate for the mutual inductance. For a 10 MHz drive
frequency, the free space wavelength is 30m; thus, Eq.3 is valid when the coil radius is
smaller than 12cm. This condition is maintained throughout the following analysis, where
a coil radius of 1 cm is modeled unless otherwise specied.
A practical question that arises relates to the necessary size of the negative permeability
slab. While the analytical model assumed a slab innite in the transverse directions, a
physical implementation would need to be nite, and preferably as small as possible. The
perfect lens material condition in the magnetostatic limit, which presumably leads to a
perfect refocusing of the magnetic near-elds, is [
x
,
y
,
z
] = [1, 1, 1]
23
. For ideal
condition, the lens has a thickness L = D/2, and the two coils have equal distance D/4 to
either surface of the lens, being placed coaxially at the object and image planes. Note that
the perfect lens condition was originally derived without consideration to there being any
power transfer between source and receiver
23
; however, even when strong coupling persists
between the source and receiver, the same conditions remain approximately optimal
31
.
To examine the eect of a nite-sized perfect lens on the mutual inductance, we sim-
ulate a nite width slab with the real part of the permeability tensor assigned the values
[
x
,
y
,
z
] = [1, 1, 1]. A small imaginary part 0.001j is added to each of the three
8
diagonal components of the permeability tensor to facilitate numerical convergence of the
nite-element scheme with the mesh size, since the exact perfect lens condition can be
numerically unstable due to the presence of strong surface modes with arbitrarily small
wavenumbers; nite damping suppresses high-k surface modes, making it possible to achieve
accurate FEM results. Thus, the permeability used in the simulations is isotropic and has
the value = 1 0.001j. Here, we use the time dependence convention E exp(jt),
for compatibility with COMSOL Multiphysics. Strong enhancements in the mutual induc-
tive coupling have been predicted in the analytical analysis, which is also revealed in the
numerical results discussed below.
As shown in Fig.2(b), the enhancement factor is, as expected, a function of W.
increases rapidly and saturates when W D. In this sub-wavelength regime, the eld
excited along the MM lens surface is a magnetostatic surface resonance (MSR)
34
, which
provides some insight as to the mechanism of the enhancement. The source coil excites a
MSR characterized by near-elds that oscillate rapidly across the slab surface and build up
strongly on the surface but decay exponentially away from it. The surface eld pattern
creates a focus of the near-elds at the optimal focus position
19,2325,35,36
, much in the
same way that elds in the aperture of a lens can produce a focus. The important dierence
between conventional lenses and the superlens, however, is that the latter connes the elds
in the focus only transversely, forming a two-dimensional focal line aligned with the point
source, along which the elds decay roughly exponentially. One would expect that, below
a certain diameter, the superlens would lose its ability to produce a perfect, predictable
two-dimensional focus; in particular, when the width of the lens is small, the eld excited on
the edge of the lens can produce uncontrolled focusing or defocusing. In terms of the surface
resonance theory, a nite-diameter superlens has a spectrum of surface modes that is spread
over a range of permeability values, in contrast with an innite-diameter lens whose spectrum
has an eigenvalue accumulation point at = 1, and thus all MSR modes with wavenumbers
k 1/L, where L is the lens thickness, can be eciently and simultaneously excited when
= 1. Exciting surface modes with a range of wavenumbers simultaneously is a key
ingredient to perfect near-eld imaging; a slab whose width W is not much larger than its
thickness L does not support a quasi-continuous range of resonant transverse wavenumbers
for any given negative value of .
The eects of the nite size of the slab can be seen in Fig.2(b), where the enhancement
9
factor becomes larger than its asymptotic value for certain slab widths. The enhancement
factor saturates for larger widths, where any inuence of the slab termination on the surface
resonance modes becomes negligible.
While the perfect lens condition places a strict limit on the thickness of the slab, we are
not necessarily interested in creating a perfect image for the WPT application; thus, for
WPT purposes, we might expect the mutual inductance to increase as the slab thickness
is increased. Fig.2(c) shows that the eld intensity in the receiver coil indeed increases as
the thickness of the slab increases with the overall distance between the source and receiver
coils xed at D = 0.5 m. The slab material properties remain the same in this study:
= 1 0.001j, = 1 . Not surprisingly, these calculations show that the closer the coils
can be placed to the slab-where the local magnetic elds are very large-the greater will be
their coupling. The distance dependence is rather sharp, since the surface modes from the
slab decay exponentially away from the slab. From the other point of view, the eective
distance between the two coils is D 2L in the presence of isotropic MM lens, comparing
to D in the absence of the MM lens. Therefore mutual inductance is much larger when the
MM lens is thicker since the mutual inductance is proportional to the inverse of distance
cubed as discussed before. The results are shown in Fig.2(c).
In the expected realizations of the negative- lens, metamaterials that exhibit articial
magnetism will likely be employed, as they have been used in recent WPT experiments
7
.
Because these metamaterials are based on inductive elements, such as split ring resonators,
they have generally signicant loss tangents related to the Ohmic losses in the constituent
current coils. The issue of loss is of particular concern for WPT applications, since the losses
represent reduced overall transfer eciency. To investigate the impact of losses in the slab,
we assume an isotropic, lossy medium of the form
e
= 1(1 + j) and = 1, where is
the magnetic loss tangent of the slab which is the ratio between the imaginary and real part
of the eective permeability. The radius for both the source and receiver coil is 0.01 m; and
the slab width is 2 m; both dimensions chosen such that all nite size eects discussed above
are negligible. The thickness of the slab correspond to the perfect lens condition: half of the
separation distance, or L = D/2 = 0.25 m . Results from analytical theory and numerical
model are shown in Fig.2(d).
Excellent agreement is again found between the analytical and numerical models, which
both show that even with a loss tangent on the order of unity, the mutual inductance is
10
still doubled. The enhancement factor has a magnitude that varies inversely with respect
to that of the magnetic loss tangent . In practice, a negative permeability metamaterials
can reach a loss tangent on the order of 0.1 or less when 1
37
. In this case, the
mutual inductance enhancement ratio can reach values in the range 10 to 100, as shown
in Fig.2(d).
IV. ANISOTROPIC, NEGATIVE-DEFINITE PERMEABILITY LENSES
For a slab made of isotropic negative permeability, the ideal thickness corresponds to D/2
, for which the source and receiver coil are placed at the object and image planes next to
the slab surfaces. The slab thus occupies half of the space between the source and receiver,
making it somewhat bulky and obtrusive. A means of reducing the thickness prole of the
slab, without changing the eld structure in any manner, can by determined by applying
techniques from transformation optics
38
.In particular, a compression transformation along
the optical axis (the direction normal to the slab or the z-direction in Fig.1(a,b,c)) results
in a material whose permeability is constant everywhere, but is anisotropic.
0 2 4 6 8
0
200
400
600
W/L0

0.5 1 1.5
50
100
200
500
1000
2000
L/L0

Numerical Result
Analytical Result
10
4
10
3
10
2
0.4
0.6
0.8
1

0
R/

0.2 0.4 0.6 0.8 1

50
100
500
1000
2000
Magnetic loss tangent ( )

Numerical Result
Analytical Result
R
w
L
(a)
(b)
(c) (d)
FIG. 2. (color online)(a) The ratio of simulated L
21
to the theoretical L
21
calculated in the
point-dipole limit, as a function of coil radius R normalized to the wavelength (coil retardation
parameter); (b) mutual inductance enhancement as a function of the MM lens width W; (c) mutual
inductance enhancement versus lens thickness for a xed inter-coil distance, where L
0
= D/2 is
the optimum lens thickness; (d) magnetic loss eect on mutual inductance enhancement.
11
As before, we consider a conguration of two coils separated by a distance D = 0.5
m, with the slab placed in between the two coils. We introduce the anisotropy factor
a to indicate the spatial compression factor. Based on transformation optics theory, the
permeability tensor required to implement a reduction in the slab prole from L to L/a
is [
x
,
y
,
z
] = [a, a, 1/a] in order to reduce the lens thickness from L
0
to L
0
/a
38
.
We performed a parametric study, where a sweeps from 0.6 to 2 and the lens thickness
L = D/(1 +a) varies from 0.625D to 0.33D. A constant imaginary part Im() = 0.001 has
been added to all three components of the slab permeability. The parametric study result
is shown in Fig.3(a).
Fig.3(a) shows that the enhancement factor is large when the anisotropy factor a is small,
in which case the receiver coil is closer to the slab surface. When the receiver is closer to
the slab, a stronger local eld exists at the receiver coil due to the surface mode. Field plots
of the slab surface mode when a = 0.6, 1, 2 are shown in Fig.3(b, c, d). As the anisotropy
factor a increases, the lens thickness decreases, and both the receiver coil and source coil
eectively move away from the slab surface. Even though the enhancement factor is reduced
for larger values of a, the overall enhancement is still large, 220 at a = 2, two orders
of magnitude larger as compared with the coupling between two coils in the absence of the
lens.
0.5 1 1.5 2
200
500
1000
2000
Anisotropy Factor: a

Numerical Result
Analytical Result
a=0.6
a=1
a=2
(a) (d)
(c)
(b)
FIG. 3. (color online)(a) Mutual coupling enhancement factor dependence on the anisotropy factor
(same as lens compression factor); System conguration and magnetic eld prole when (b) a = 0.6,
(c) a = 1, (d) a = 2.
The change in the enhancement factor is initially surprising, since the coordinate trans-
formation should preserve all properties of the eld distribution perfectly. That is, the
compressed slab with anisotropic medium should perform as well as the isotropic slab, with
12
the enhancement factor remaining constant for any value of the anisotropy factor a. The
basic transformation optics theory, however, is geometrical in nature and does not imme-
diately apply to complex material parameters. We thus suspect that the reason for the
variation in the mutual coupling enhancement relative to the transformation optics predic-
tion lies in the loss tangent applied to the anisotropic medium. In order to preserve the
same performance as isotropic lens, the loss in the anisotropic medium must be scaled ap-
propriately. We investigated scaling the loss tangent of the slab in several dierent ways:
= 0.1, 0.1/a, 0.1/a
2
. Results for the dierent scaling are shown in Fig.4(a). As shown, an
anisotropic medium with = 0.1/a
2
yields the expected at (a-independent) enhancement
factor .
Having identied the proper scaling for the loss tangent with respect to the anisotropy
factor, we study the impact of losses on the anisotropic slab. In the following, all components
of permeability tensor share the same loss tangent . For = 1/a
2
the enhancement factor
reduces from a value around 50 to a unit value. Both the analytical and numerical result
is shown in Fig.4(b). In the practical case where 0.025, the enhancement still achieves
around 50 even when the slab is only quarter of the transfer distance.
0.5 1 1.5 2
0
10
20
30
40
50
60
Anisotropy factor: a

2
=0.1/a
2
Analytical
=1/a
2
Analytical
=0.1/a
2
Numerical
=1/a Numerical
0.5 1 1.5 2
50
100
150
200
Anisotropy factor: a

=0.1 Analytical
=0.1/a Analytical
=0.1/a
2
=0.1 Numerical
=0.1/a Numerical
=0.1/a
2
Analytical
Numerical
(a) (b)
FIG. 4. (color online) (a)Enhancement factor depends on the anisotropic MM lenss material
loss tangent = 0.1, 0.1/a, 0.1/a
2
; (b) Mutual enhancement reduction while increase the material
loss tangent from = 0.1/a
2
to = 1/a
2
The results in the above study show that material loss needs to be well controlled in the
anisotropic lens, whenever its thickness is less than the thickness of the canonical, isotropic
lens (i.e. D > L/2 and a > 1). In order to achieve the same magnitude of coupling with
a lens with a > 1, the magnetic loss tangent needs to decrease roughly as 1/a
2
. In other
words, there is a trade-o between increasing the transfer distance of the lens relative to its
thickness, and decreasing resistive loss in the metamaterial.
13
V. INDEFINITE-PERMEABILITY LENSES
In the previous section, we have analyzed the performance of anisotropic lens, which en-
ables reduction of the MM lenss thickness in the power transfer direction. Now we study the
system in which the MM lens is formed by an indenite-permeability medium (or indenite
medium for short), where one of the three principal values of the permeability tensor has
dierent sign from the other two
3941
. The motivation to apply indenite medium is to sim-
plify the MM lens conguration in reality. By eliminating one (or two) negative component
in the permeability tensor, one (respectively, two) resonant unit can be removed from each
array element in the periodic articial composite
37
. This reduces the fabrication complexity,
and even allows one to reduce the space between magnetic resonators and stack them more
densely for enhanced magnetic oscillator strength and better magnetic loss tangents in the
negative- band
37
.Equivalent performance of WPT system with anisotropic MM lens has
been veried experimentally comparing to the WPT system with isotropic MM lens
7
.
An important distinction of the indenite-medium lens is that it supports propagation
of waves with wavenumbers greatly exceeding the free space wavenumber k
0
= /c, which
is a consequence of the hyperbolic dispersion relation in an indenite medium
4250
. Thus,
even in the near eld, spatial Fourier components that are evanescent in free space become
propagating waves inside the indenite medium. In our specic study, the anisotropic ma-
terial properties for the MM lens is:
z
= b 0.1j,
x
=
y
= 1 and = 1. And the lens
has a thickness L = D/2, with nite width in the numerical model, and innite width in
analytical model. The analytical result predicts the enhancement ratio is less than 10 (the
solid blue curve in Fig.5). Two indenite MM lens with dierent lens width W = 3.6m and
W = 4m are studied numerically. The numerical simulations, which utilize nite slab width
W, reveal two superimposed oscillations in the vs b curve.
The short oscillations in the numerical result are due to the Fabry-Perot resonances(FPR),
which result from the nite width W of the MM lens and are enabled by the above-mentioned
hyperbolic dispersion property of the indenite medium. Slab termination at a nite radius
(W/2) creates additional reecting planes and possibility for standing wave, which are not
accounted for in the analytical model that assumes an innitely wide slab.
To make a better comparison with the idealized analytical model, we have applied a curve
tting technique that eliminates these short oscillations. The tting curve (dashed green
14
curve in Fig.5) based on the numerical result matches well with the analytical prediction.
In the tted curve, as well as the analytical calculation, there is yet another, longer-period
oscillation as a function of b = Re(
z
); since those are captured by the idealized model,
but are not present in the negative-denite lens case, we must attribute this longer-period
oscillation to the FPR associated with the thickness D of the lens. Since D W in all cases
considered, FPR of the lens thickness D are indeed expected to have a substantially longer
period than FPR of the lens width W, which further conrms our explanation of both types
of oscillations.
0 1 2 3 4
0
2
4
6
8
10
b

Analytical
Numerical: lens width 4m
Numerical: lens width 3.6m
Oscillation-removing fit for W=3.6m
FIG. 5. (color online)Indenite-permeability lens: coupling enhancement as a function of b =
Re(
z
).
Finally, Finally, we summarize typical enhancement ratios that can be obtained with
isotropic, anisotropic all-negative lens (which is based on the transformation optics ideas),
and two indenite-medium lenses. The result is shown in Fig.6. In this summary, both
isotropic and indenite medium lens share the same loss tangent = 0.1 and lens thickness
L = D/2 = 0.25m. The anisotropic denite-negative permeability lens has loss tangent
= 0.1/a
2
and lens thickness L = D/4 = 0.125m since the anisotropy factor a = 2. In the
indenite medium lens, the oscillation in results from FPR modes associate with the nite
dimension of the MM lens.
To summarize the results in Fig.6, we nd that with a suciently small metamaterial
loss, the inductive coupling enhancement provided by the negative-denite anisotropic per-
15
meability lens can be as substantial good as with an isotropic negative permeability lens
of a substantially larger thickness. This result shows that anisotropic MM lenses can be
potentially applied to reducing the size, volume and mass of the metamaterial component
in the MM-enhanced wireless power systems.
With the indenite permeability lens, optimum mutual coupling enhancement can be
achieved by constructively rebuilding the eld in the receiver coil and manipulating the
MSR mode on the lens surface with the boundary eect. In particular, when the lens has
width equal to 0.15m, and the lens with
z
= 2 0.2j, the mutual coupling enhancement
in the indenite lens outperforms that of the isotropic and anisotropic MM lens when all
three lenses share the same width. The results shown here predict that, in some cases, the
indenite permeability lens can outperform isotropic or anisotropic negative-denite lenses
when designed and optimized to take advantage of the eects related to the nite slab size
and boundary reections.
0 0.2 0.4 0.6 0.8 1
0
10
20
30
40
50
Lens radius: W/2

Indefinite lens: b=1
Indefinite lens: b=2
Isotropic lens
Anisotropic
all-negative lens
FIG. 6. (color online)The summary of enhancement ratio obtained with isotropic MM lens,
anisotropic MM lens and indenite lens when the lens radius changes. Isotropic lens has
= 10.1j and L = D/2 in which D = 0.5m; Anisotropic lens has [
x
,
x
,
x
] = [20.05j, 2
0.05j, 0.5 0.0125j] and L = D/4; Indenite lens with b = 1 has [
x
,
y
,
z
] = [1 0.1j, 1, 1]
and L = D/2; Indenite lens with b = 2 has [
x
,
y
,
z
] = [2 0.2j, 1, 1] and L = D/2.
16
VI. CONCLUSIONS
We have analyzed the mutual inductive coupling enhancement by virtue of MM lenses
added to the resonant near-eld wireless power transfer systems. A few practical issues
have been addressed here, including nite dimension of both the coil and MM lens, and
the application of the anisotropic negative-denite and indenite permeability lenses. In
our results, the numerical result matches the analytical prediction with excellent precision
whenever the numerical model operates within the assumptions of the analytical model. We
nd that the inductive coupling enhancement can be substantial even when the MM lens has
signicant resistive loss. The application of anisotropic, negative-denite permeability lenses
requires ne control over the material loss, but enables the reduction of the metamaterial
layer thickness. To make use of an indenite permeability lens, the lens dimensions and
permeability magnitude need to be carefully optimized in order to achieve the maximum
possible enhancement; in certain regimes, the indenite lens can perform as well as the
denite-negative lens or even outperform the latter.
ACKNOWLEDGMENTS
The authors are thankful to Matt Reynolds (Duke University) and Bingnan Wang (Mit-
subishi Electric Research Laboratories) for useful discussions on wireless power transfer.
REFERENCES
1
NOLIT, Nikola Tesla Colorado Springs Notes 1899-1900 (Nilola Tesla Museum, Beograd,
1978) p. 395.
2
H. Matsumoto, K. Hashimoto, N. Shinohara, and T. Mitani, in Proceedings of The 4th
International Conference on Solar Power from Space (2004).
3
J. C. Mankins, Space Solar Power The rst International Assessment of Space Solar Power: Opportunities,
Rep. (International Academy of Astronautics, 2011).
4
T. Deyle and M. Reynolds, 2008 IEEE International Conference on Robotics and Automation , 1036 (2008)
5
B. L. Cannon, J. F. Hoburg, D. D. Stancil, and S. C. Goldstein, IEEE Transactions on
Power Electronics 24, 1819 (2009).
17
6
Z. N. Low, R. A. Chinga, R. Tseng, and J. Lin,
IEEE Transactions on Industrial Electronics 56, 1801 (2009).
7
B. Wang, K. H. Teo, T. Nishino, W. Yerazunis, J. Barnwell, and J. Zhang,
Applied Physics Letters 98, 254101 (2011).
8
C. Reinhold, P. Scholz, W. John, and U. Hilleringmann, Journal of Communications 2,
14 (2007).
9
X. Yu, S. Sandhu, S. Beiker, R. Sassoon, and S. Fan,
Applied Physics Letters 99, 214102 (2011).
10
M. W. Cardullo,W. L. Park III, US Patent, 3,713,148, (1973).
11
U. Zink and G. Skuro, US Patent 6,661,197 B2, (2003).
12
G. W. Bruning, US Patent 2003/0231001 A1, (2008).
13
A. K. Ramrakhyani, S. Member, and S. Mirabbasi, IEEE Transaction on Biomedical
Circuits and Systems 5, 48 (2011).
14
E. Waenschmidt and T. Staring, EPE 09. 13th European Conference on Power Electron-
ics and Applications , 1 (2009).
15
A. Kurs, A. Karalis, R. Moatt, J. D. Joannopoulos, P. Fisher, M. Soljac, and M. Soljacic,
Science 317, 83 (2007).
16
A. P. Sample, D. A. Meyer, and J. R. Smith,
IEEE Transactions on Industrial Electronics 58, 544 (2011).
17
P. Si, A. P. Hu, S. Malpas, and D. Budgett,
IEEE Transactions on Biomedical Circuits and Systems 2, 22 (2008).
18
R. Hasegawa, Materials Science and Engineering A 375-377, 90 (2004).
19
M. C. Wiltshire, J. B. Pendry, I. R. Young, D. J. Larkman, D. J. Gilderdale, and J. V.
Hajnal, Science 291, 849 (2001).
20
M. J. Freire, R. Marques, and L. Jelinek, Applied Physics Letters 93, 231108 (2008).
21
E. Shamonina and L. Solymar, Journal of Physics D: Applied Physics 37, 362 (2004).
22
V. Veselago, Physics-Uspekhi 10, 509 (1968).
23
J. Pendry, Physical review letters 85, 3966 (2000).
24
S. A. Cummer, Applied Physics Letters 82, 1503 (2003).
25
R. Merlin, Science 317, 927 (2007).
26
A. Grbic, L. Jiang, and R. Merlin, Science 320, 511 (2008).
18
27
D. R. Smith, D. Schurig, M. Rosenbluth, S. Schultz, S. A. Ramakrishna, and J. B. Pendry,
Applied Physics Letters 82, 1506 (2003).
28
N. Fang, Z. Liu, T. Yen, and X. Zhang, Applied Physics A 80, 1315 (2005).
29
T. Cui, Q. Cheng, W. Lu, Q. Jiang, and J. Kong, Physical Review B 71, 1 (2005).
30
K. H. Teo, D. Huang, and B. Wang, US Patent 2011/0133564 A1, (2011).
31
Y. Urzhumov and D. R. Smith, Physical Review B 83, 31 (2011).
32
H. A. Ghali and H. A. Rahman, in Proceedings of the COMSOL conference 2009 (Milan,
2009).
33
F. W. Grover, Inductance Calculations: Working Formulas and Tables (Instrumentation
Systems &, 1982) p. 300.
34
Y. Urzhumov, W. Chen, C. Bingham, W. Padilla, and D. R. Smith, Phys. Rev. B 85,
054430 (2003).
35
G. Shvets, Proceedings of SPIE 5221, 124 (2003).
36
M. J. Freire, L. Jelinek, R. Marques, and M. Lapine,
Journal of magnetic resonance (San Diego, Calif. : 1997) 203, 81 (2010).
37
W. C. Chen, C. M. Bingham, K. M. Mak, N. W. Caira, and W. J. Padilla,
Extremely Sub-wavelength Planar Magnetic Metamaterials, (2011), submitted,
arXiv:arXiv:1111.2024v1.
38
D. A. Roberts, N. Kundtz, and D. R. Smith, Optics express 17, 16535 (2009).
39
D. Schurig and D. R. Smith, New Journal of Physics 162, 0 (2005).
40
H. Lee, Z. Liu, Y. Xiong, C. Sun, and X. Zhang, Optics express 15, 15886 (2007).
41
J. Yao, X. Yang, X. Yin, G. Bartal, and X. Zhang,
Proceedings of the National Academy of Sciences of the United States of America 108, 11327 (2011).
42
D. R. Smith and D. Schurig, Physical Review Letters 90, 077405 (2003).
43
A. Al` u and N. Engheta, Journal of Applied Physics 97 (2005).
44
D. Schurig and D. R. Smith, Applied Physics Letters 82, 2215 (2003).
45
K. G. Balmain, A. A. E. L uttgen, and P. C. Kremer, IEEE Antenna and Wireless Prop-
agation Letters 1, 146 (2002).
46
Z. Jacob, L. V. Alekseyev, and E. Narimanov, Optics Express 14, 8247 (2006).
47
D. R. Smith, P. Kolinko, and D. Schurig, Journal of the Optical Society of America B 21, 1032 (2004).
48
D. R. Smith, D. Schurig, J. J. Mock, P. Kolinko, and P. Rye,
Applied Physics Letters 84, 2244 (2004).
19
49
O. F. Siddiqui and G. V. Eleftheriades, Journal of the Franklin Institute 348, 1285 (2011).
50
Q. Cheng, R. Liu, J. Mock, T. Cui, and D. R. Smith, Physical Review B 78, 121102 (2008)
.
20